| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide a device capable of rasing generalization capability of a classifier in a learning device of pattern classifier by LGM-MCE learning. <P>SOLUTION: An expression (1) is used to define an error classification scale value D<SB POS="POST">y</SB>(x;Λ) for measuring a degree of error classification of a sample x belonging to a class C<SB POS="POST">y</SB>. An expression, g<SB POS="POST">y</SB>(x;Λ) based on ψ>0, shows a discrimination function for a belonging degree of x to C<SB POS="POST">y</SB>. A learning device executes: a step for obtaining the error classification scale values of the sample belonging to C<SB POS="POST">y</SB>; a step for, with setting a real probability distribution generating them as a Parzen distribution of a window width h<SB POS="POST">y</SB>centering on each of the error classification scale values, evaluating likelihood of the distribution with a function of h<SB POS="POST">y</SB>; a step for estimating the likelihood through cross-validation most likelihood estimation; a step for using an expression, α<SB POS="POST">y</SB>=4/((2π)<SP POS="POST">1/2</SP>*h<SB POS="POST">y</SB>), to calculate an optimal value α<SB POS="POST">y</SB>of lost smoothness, with respect to h<SB POS="POST">y</SB>that provides a most likelihood distribution; and a step for adjusting a learning parameter Λ so as to minimize an empirical average loss that is a function of α<SB POS="POST">y</SB>. <P>COPYRIGHT: (C)2012,JPO&INPIT |
